Benefit of dexamethasone compared with prednisolone for childhood acute lymphoblastic leukaemia: results of the UK Medical Research Council ALL97 randomized trial


Dr Mitchell, Paediatric Haematology/Oncology, John Radcliffe Hospital, Oxford OX3 9DU, UK.


Corticosteroids are an essential component of treatment for acute lymphoblastic leukaemia (ALL). Prednisolone is the most commonly used steroid, particularly in the maintenance phase of therapy. There is increasing evidence that, even in equipotent dosage for glucocorticoid effect, dexamethasone has enhanced lymphoblast cytotoxicity and penetration of the central nervous system (CNS) compared with prednisolone. Substitution of dexamethasone for prednisolone in the treatment of ALL might, therefore, result in improved event-free and overall survival. Children with newly diagnosed ALL were randomly assigned to receive either dexamethasone or prednisolone in the induction, consolidation (all received dexamethasone in intensification) and continuation phases of treatment. Among 1603 eligible randomized patients, those receiving dexamethasone had half the risk of isolated CNS relapse (P = 0·0007). Event-free survival was significantly improved with dexamethasone (84·2% vs. 75·6% at 5 years; P = 0·01), with no evidence of differing effects in any subgroup of patients. The use of 6·5 mg/m2 dexamethasone throughout treatment for ALL led to a significant decrease in the risk of relapse for all risk-groups of patients and, despite the increased toxicity, should now be regarded as part of standard therapy for childhood ALL.

The probability of long-term event-free survival (EFS) for childhood acute lymphoblastic leukaemia (ALL) has improved to over 75% with modern intensive multidrug chemotherapy. Two of the longest recognized and important drugs in ALL therapy are the steroid, prednisolone, used to induce remission and as part of continuing therapy and the Thiopurine, 6-mercaptopurine (6-MP), which is used throughout the continuation phase (Childhood ALL Collaborative Group, 1996). Improved understanding of the pharmacology and pharmacokinetics of these agents has given rise to the suggestion that the analogues 6-thioguanine (6-TG) and dexamethasone might be more effective than 6-MP and prednisolone, respectively. The Medical Research Council (MRC) protocol ALL 97 sought to establish whether this notion might be correct. This report focuses on the steroid randomized question: a future paper will explore results of the randomized trial between 6-TG and 6-MP.

Although in terms of conventional glucocorticoid activity, dexamethasone is only 6·25 times more potent than prednisolone; the MTT (3-2, 5-diphenyltetrazolium bromide) assay suggests a 16-fold gain in potency against lymphoblasts (Kaspers et al, 1996). In addition, it has been suggested that dexamethasone has better penetration of the central nervous system (CNS) (Balis et al, 1987) and there is clinical evidence that substitution of prednisolone by dexamethasone results in a lower incidence of meningeal leukaemia (Jones et al, 1991). For example, the Dutch ALL study VI substituted dexamethasone for prednisolone with major gains in EFS compared with their historical experience (3-year EFS 66% for the comparable study V vs. 80% for patients entered into study VI) (Veerman et al, 1996). The Cancer and Leukaemia Group B (CALGB), also in a historical comparison, found that children assigned to dexamethasone had a lower CNS relapse rate than those assigned to prednisolone, although the EFS for the two groups was similar (Holland & Glidewell, 1972).

Initially the ALL97 treatment regimens were based on those used in the immediately preceding trial, UKALL XI (Hann et al, 2001). After the comparative analysis for trial outcome during the period 1980–97 in the UK and in 11 other countries was reported (Schrappe et al, 2000), it was recognized that other regimens were giving better disease control than UKALL XI (Eden et al, 2000). The type of intensification modules used and stratification of treatment by risk factors, including initial response, was amended in 1999 to resemble more closely the US Children's Cancer Study Group (CCG) trial 1922 protocol (Bostrom et al, 2003). The randomized questions were continued into the modified trial.



ALL 97 opened for recruitment in April 1997 for children with standard and high risk ALL, aged 1–18 years. All centres entering patients received Local Ethical Committee approval for trial participation. After obtaining informed consent from the parents and, where appropriate, the patient, patients were randomly assigned to either prednisolone 40 mg/m2/d p.o. in two divided doses or dexamethasone 6·5 mg/m2/d p.o. in two divided doses. The steroid to which the patient was randomized was the steroid to be received during induction therapy and the 5-day pulses throughout interim maintenance and continuing therapy. The second randomization was between 6-mercaptopurine 75 mg/m2 daily in a single dose or 6-thioguanine 40 mg/m2 daily in a single dose during continuing therapy. In the first part of the trial, patients at very high risk according to the Oxford hazard score (Chessells et al, 1995a), based on age, white blood cell count (WBC), and gender, were not randomized, but treated on a more intensive protocol (ALL-HR1) (unpublished observations).

The overall treatment template changed in several more minor ways during the course of the study. The initial treatment regimen used for the period 1997–99 was, apart from the randomized components, identical to that used in UKALL XI from 1992 to 1997 (Hann et al, 2001). The induction period included weekly intravenous vincristine 1·5 mg/m2 (with a maximum dose of 2 mg), daily oral steroid as randomized and Erwinia asparaginase (Erwinase), at a dose of 6000 μg/m2 per dose for nine doses given on a Monday, Wednesday and Friday. No induction anthracycline was used. Short intensification blocks were then given at weeks 5 and 20; patients continued to be randomized to receive or not a third intensification block at week 35 as previously described (Hann et al, 2001). These details are summarized in Table I. In May 1998, a clear benefit was seen (Hann et al, 2000) for three rather than two intensifications and all subsequent patients received three intensification modules. At this time it was also recommended that all patients who had not reached week 35 should receive a third intensification phase. In April 1998, because of pharmacokinetic data indicating that the dose of Erwinase used was sub-optimal, the administration was changed to 12 doses of 6000 μg/m2 given on alternate days (unpublished observations).

Table I.  ALL97 treatment regimens.
InductionFirst IBCNS-directed treatmentSecond IBInterim CTThird IBCT
  1. 1Randomized in ALL97. All patients received prednisolone in UKALL XI.

  2. 2WCC < 50 × 109/l received IT MTX alone in ALL97.

  3. 3Randomized in ALL97. All patients received mercaptopurine in UKALL XI.

  4. 4Or dexamethasone in ALL97 according to induction randomization.

  5. 5105 weeks in ALL97.

  6. IB, intensification block; CNS, central nervous system; CT, continuing therapy; VCR, vincristine; L-ASP, Erwinia–asparaginase; IT MTX, intrathecal methotrexate; HD MTX, high dose intravenous methotrexate; p.o., oral; i.v., intravenous; s.c., sub-cutaneous; i.m., intramuscular.

VCR 1·5 mg/m2 i.v. days 1, 7, 14 and 21
Prednisolone 40 mg/m2 p.o. days 1–28 or
6·5 mg/m2/day p.o. days 1–28
6000 μg/m2 s.c./i.m. 3/week × nine doses
12·5 mg/m2 days 1 and 8
VCR 1·5 mg/m2 i.v. day 1
Prednisolone 40 mg/m2 p.o. × 7 days
100 mg/m2 i.v. × 5 days
100 mg/m2 i.v. days 1–5
45 mg/m2 days 1 and 2
80 mg/m2 p.o. days 1–5
12·5 mg/m2 day 1
WBC < 50 × 109/l
12·5 mg/m2 weekly weeks 9–12 or
6 or 8 g/m2 weeks 6, 8 and 10
12 hourly × 5 days plus IT MTX
weeks 5, 6, 8 and 10
WBC ≥ 50 × 109/l
HD MTX as above
24 Gy Cranial Radiotherapy
in 15 fractions of 1·6 Gy each
weeks 9–12 (except 1–2 years age who were allocated HD-MTX)
Interim CT
Mercaptopurine 75 mg/m2
or Thioguanine 40 mg/m2 p.o. daily
Methotrexate 20 mg/m2 p.o. weekly except during CNS-directed treatment
VCR 1·5 mg/m2 i.v. and
Prednisolone4 40 mg/m2 p.o. daily × 5 days every 4 weeks
VCR 1·5 mg/m2
i.v. day 1
Prednisolone 40 mg/m2
p.o. × 7 days
100 mg/m2 i.v. × 5 days
100 mg/m2 i.v.
12 hourly × 5 days
45 mg/m2 days 1 and 2
12·5 mg/m2 day 1
Thioguanine 80 mg/m2 p.o. days 1–5
Same as weeks 8–19Dexamethasone
10 mg/m2 p.o. daily for 10 days, then 4 day taper
VCR 1·5 mg/m2 i.v. days 1, 7, 14 and 21
6000 μg/m2 s.c./i.m. 3/week × nine doses
12·5 mg/m2 days 1 and 28
600 mg/m2 days 28 and 42
75 mg/m2 days 28–31, 35–38, 42–45 and 49–52
60 mg/m2 p.o. days 28–56
3 monthly IT MTX
12·5 mg/m2
Same as Interim
Weeks 1–45 + 2 for recovery8–1920 + 2 for recovery22–3435–4242–1005

In November 1999, the basic therapy template and randomized questions were retained but the intensification pulses were changed from 5-d, short course intensive treatment, to type sustained, but less intensive, modules of the type pioneered by the Berlin Frankfurt Munster (BFM) group (Schrappe et al, 2000). This phase of the trial was designated ALL97/99 and is summarized in Table II.

Table II.  ALL97/99 treatment regimens.
InductionConsolidationInterim maintenance 1Delayed intensification 1Interim maintenance 2Delayed intensification 2Continuing therapy
Regimen A
VCR 1·5 mg/m2 days 1, 8, 15, 22 and 29
40 mg/m2/day p.o. days 1–29 or
6·5 mg/m2/day p.o. days 1–29
l-asparaginase (Elspar)
6000 μg/m2 s.c./i.m., 3/week × nine doses
If switching to regimen C,
Daunorubicin 45 mg/m2 i.v., days 15 and 22
Steroids: Taper steroids over 7 days to zero, days 1–7
VCR 1·5 mg/m2 i.v. on day 1 (week 5)
6-mercaptopurine 75 mg/m2/day p.o., or
6-thioguanine 40 mg/m2/day p.o. daily on days 1–29 (weeks 5–8)
Steroid as randomized for 5 days starting on days 1 and 29
VCR 1·5 mg/m2 days 1 and 29
Thiopurine as randomized for days 1–49
Methotrexate 20 mg/m2 p.o. weekly on days 8, 15, 22, 36, 43, and 50
Dexamethasone 10 mg/m2 p.o. for 7 days starting on days 1 and 15
VCR 1·5 mg/m2 days 1, 8 and 15
Doxorubicin 25 mg/m2 on days 1, 8, 15
l-Asparaginase (Elspar) 6000 μg for six doses stating on day 4
Cyclophosphamide 1 g/m2 i.v. on day 29
Thioguanine 60 mg/m2 days 29–42
Cytarabine 75 mg/m2 i.v., days 30–33 and 37–40
Steroid as randomized for 5 days starting on days 1 and 29
VCR 1·5 mg/m2 days 1 and 29
Thiopurine as randomized for days 1–49
Methotrexate 20 mg/m2 p.o. weekly on days 8, 15, 22, 36, 43 and 50
Dexamethasone 10 mg/m2 p.o. for 7 days starting on days 1 and 15
VCR 1·5 mg/m2 days 1, 8 and 15
Doxorubicin 25 mg/m2 on days 1, 8, 15
l-Asparaginase (Elspar) 6000 μg for six doses stating on day 4
Cyclophosphamide 1 g/m2 i.v. on day 29
Thioguanine 60 mg/m2 days 29–42
Cytarabine 75 mg/m2 i.v., days 30–33 and 37–40
12-week cycles consisting of:
Steroid as randomized for 5 days starting on days 1, 29 and 57
VCR 1·5 mg/m2 days 1 and 29
Thiopurine as randomized for days 1, 29, 57
Thiopurine as randomized
Methotrexate 20 mg/m2 p.o. weekly on days 8, 15, 22, 36, 43, and 50
Regimen B
As per regimen A plus
Daunorubicin 25 mg/m2 i.v. on days 1, 8, 15 and 22
No change in induction if switching to regimen C
Standard BFM consolidation
Cyclophosphamide 1 g/m2 i.v. on days 1 and 15
Cytarabine 75 mg/m2 i.v., days 2–5, 9–12, 16–19 and 23–26
6-mercaptopurine, 60 mg/m2 days1–29
As per regimen AAs per regimen AAs per regimen AAs per regimen AAs per regimen A
Regimen C
Induction as per modification to regimen A above, or else as per regimen B aboveAugmented BFM consolidation
Cyclophosphamide 1 g/m2 i.v. on days 1 and 29
Cytarabine 75 mg/m2 i.v., days 2–5, 9–12, 30–33 and 37–40
6-mercaptopurine, 60 mg/m2 days 1–15 and 29–42
VCR 1·5 mg/m2 i.v. on days 15, 22, 43, 50
PEG asparaginase 2500 μg/m2 days 15 and 43
Capizzi I
VCR 1·5 mg/m2 i.v. on days 1, 11, 21, 31 and 41
Methotrexate 100 mg/m2 increasing by 50 mg/m2 every 10 days as permitted by toxicity
PEG asparaginase 2500 mug/m2 days 2 and 22
As per regimen ACapizzi II
VCR 1·5 mg/m2 i.v. on days 1, 11, 21, 31 and 41
Methotrexate 100 mg/m2 increasing by 50 mg/m2 every 10 days as permitted by toxicity
PEG asparaginase 2500 μg/m2 days 2 and 22
As per regimen AAs per regimen A
Intrathecal therapy
Cytarabine day 1 (week 1) only. Dose by age:  < 2 years: 30 mg; 2 years: 50 mg; ≥ 3 years: 70 mg
Methotrexate day 8 (week 2)
Dose by age: < 2 years: 8 mg; 2 years: 10 mg; ≥ 3 years: 12 mg
Methotrexate days 1, 8, 15, 22, dose as inductionMethotrexate on days 1 and 31, doses as inductionMethotrexate on days 1 29 and 36, doses as inductionMethotrexate on days 1 and 31, doses as inductionMethotrexate on days 1 29 and 36, doses as inductionMethotrexate on days 1 of each cycle, doses as induction

During the latter phase, initial stratification based on age and presenting white cell count was used to place patients on regimen A (standard risk) or regimen B (high risk). The speed of response was assessed at day 8 for regimen B and day 15 for regimen A. An M3 marrow (25% blasts or greater) at day 8 on regimen B or at day 15 on regimen A led to transfer to the very high-risk regimen, C; these patients also continued to receive the type of steroid as previously randomized. Patients with near-haploidy, Philadelphia-positive ALL or MLL gene rearrangements were transferred to regimen C as soon as those cytogenetic abnormalities were identified. Patients who failed to remit on regimens A and B by day 28 were transferred to regimen C and considered for bone marrow transplantation. In addition, duration of continuing therapy was increased for boys so that the total length of therapy was 3 years whilst girls continued to receive 2 years.

In April 2001, the form of asparaginase used was changed from the Erwinia chrysanthemi-derived Erwinase to the E. coli-derived Elspar (Merck Sharpe & Dohme, Hoddesden, Hertfordshire, UK) at a dosage of 6000 μg/m2 intramuscularly per dose for nine doses, given on Monday, Wednesday and Friday. This change in regimen was based on comparative pharmacological data, especially the near doubling of the half-life of Elspar compared with that of Erwinase. (Dr V.I. Avramis, personal communication). The change also resulted in the trial regimens recapitulating those used in the CCG 1922 study (Bostrom et al, 2003).

In ALL97, CNS-directed therapy for all patients was with 16 doses of intrathecal methotrexate with a dosage that was based on age. This treatment was amended in ALL 97/99 to: regimen A, girls: 19 doses, boys: 23 doses, regimen B and C, girls: 22 doses, boys: 26 doses. Patients with CNS disease present at diagnosis received additional intrathecal medication during induction until the CSF was clear, and 24Gy of cranial irradiation during the consolidation phase.

The major toxicity assessed was treatment-related mortality during induction and/or remission. Other toxicities recorded were infections during each phase of therapy and National Cancer Institutes (NCI) grade III and IV toxicity for any of the following potential steroid sequelae: acute behavioural disturbance, hypertension, diabetes mellitus, myopathy, avascular necrosis of joints, osteopenia, and any other recognized steroid toxicity (for example, excessive weight gain, acute gastrointestinal upset, hepatomegaly).

In 2002, the Data Monitoring Committee recommended closure of the trial because of the observed benefit of dexamethasone over prednisolone in ALL97/99, and because the data emerging from this trial were consistent with the results supplied in confidence by the US CCG, which were subsequently reported (Bostrom et al, 2003). It was recommended that all patients still on therapy should be converted to dexamethasone for the remainder of their continuing therapy.

Statistical analyses

Randomization was by telephone call to a central office where patient details were recorded, and then treatment allocated by computer using minimization to balance treatments over gender, age, WBC and other treatment allocation groups. The EFS difference suggested by the Dutch historical comparison (Veerman et al, 1996) was about 15%, but in case this estimate was inflated, a sufficient number of patients were to be recruited to give the statistical power to detect a smaller difference. The target was set at 1800 randomized patients, which would give over 99% power to detect a 10% difference and 80% power to detect a 6% difference, assuming a baseline EFS of 70% at 4 years. The results of interim analyses were reported annually to the independent MRC Leukaemia Data Monitoring Committee.

The main analyses were of EFS, with event defined as time to relapse or death, and of survival, using the log rank method for comparison between groups. Secondary outcomes were death in induction, remission death, isolated CNS relapse, CNS relapse combined with relapse at another site, and non-CNS relapse, defined as relapse without CNS involvement. Analyses were all of first event, censored at events other than the one of interest. All treatment comparisons are by intended treatment, regardless of whether it was received or not.

As well as Kaplan–Meier plots of survival by treatment allocation, odds-ratio plots are used to show the relative effect of steroid type within subgroups (Antiplatelet Trialists’ Collaboration, 1988). Within each subgroup, the observed minus expected (O − E) number of events and its variance (V) are given. The odds ratio, calculated from these data, is shown as a filled box, with a line indicating the 95% confidence interval (CI). The overall effect after stratification by each subgroup variable is shown as a diamond whose width indicates the 95% CI of the overall result. The results of tests for heterogeneity, or trend in the case of WBC, are displayed to indicate any evidence of a different effect in any subgroup.

Current analyses were to the annual follow-up of 31st October 2004, with a median follow-up from diagnosis of 4 years and 11 months.

Toxicity was analysed with the SAS statistical package, using chi-square and Cochrane–Mantel–Haenszel tests, and by logistic regression. P-values < 0·05 were considered significant.


A total of 1948 patients had been registered by the time the trial was closed in June 2002. Thirteen patients were excluded for misdiagnosis (seven acute myeloid leukaemia, two mature B-cell ALL, three non-Hodgkin's lymphoma, and one isolated CNS disease with <5% involvement in the bone marrow). Nine patients were lost to follow-up, six of them before 5 years had elapsed, with five of them randomized for type of steroid. A total of 1621 patients were randomized for steroid but 18 were subsequently excluded from analyses because they were found during induction to be high risk (HR) and were treated on the HR1 regimen, thus leaving 1603 randomized and appropriately treated patients (see Fig 1).

Figure 1.

Consolidated Standards for Reporting of Trials (CONSORT) diagram.

Compared with the previous trial, UKALL XI, EFS had improved (Fig 2), although overall survival (OS) was similar. For patients treated in the first phase of the trial (ALL97), including the 151 very high risk patients who were treated on the ALLHR1 protocol, the EFS at 5 years was 74·1% (95% CI = 71·4–76·8%) compared with 63·1% (95% CI = 60·9–65·3%) for UKALL XI. The follow-up for the second phase of the trial (ALL99) is short at present, but the 5-year EFSof 81% (95% CI = 77·9–84·1%), suggests further improvement.

Figure 2.

Event-free survival by treatment protocol.

Table III shows the allocation of patients to each background treatment and the steroid allocations. The 1603 children who were both eligible for the main trial and were randomized for steroid form the main focus of this paper.

Table III.  ALL97/99 trial entrants and steroid randomization rates and numbers.
 Total patientsALL 97 (1997–99)ALL 99 (1999–2002)
Main trialHR1Regimen ARegimen BRegimen C*
  1. *Arm C patients came from Arms A and B as a result of: (1) identification of the cytogenetic abnormalities, near haploidy, Ph + ALL, or MLL gene rearrangements; (2) as a result of slow response on Arms A or B.

Number entered1935846151556232150
Total randomization (% of entrants)1621781 (92%)18497 (89%)199 (86%)126 (84%)
Non-randomized dexamethasone1610474
Non-randomized prednisolone29864133552620

A total of 798 patients were randomized to receive dexamethasone and 805 to receive prednisolone during induction and continuing therapy. The rate of randomization was over 90% during the first phase of the trial, and remained high, at over 84%, in all arms in the second phase. Sixteen patients/parents/doctors violated the protocol and opted for dexamethasone. The remaining 165 eligible patients who declined randomization received prednisolone. There were no obvious differences in demographic or leukaemia characteristics between those refusing randomization and those randomized. Initial characteristics by treatment allocation within trial phase were well balanced (Table IV).

Table IV.  Patient diagnostic characteristics.
 ALL97 (not HR1)ALL99Total
 M195196240236867 (54%)
 F197193173173736 (46%)
Age (years)
 <234322929124 (8%)
 2–93113063013011219 (76%)
 ≥1047518379260 (16%)
WBC (×109/l)
 <10208196198199801 (50%)
 10–1963645566248 (15%)
 20–4959716755252 (16%)
 50–9936293752154 (10%)
 ≥10026295637148 (9%)
Ph +ve or bcr-abl +ve10111022 (1%)
t(4;11)105410 (0·6%)
11q23/MLL rearrangement128415 (1%)

Outcome by treatment

There was a significant reduction in the risk of isolated central nervous system relapse for those receiving dexamethasone. The actuarial isolated CNS relapse rate at 5 years was 2·5% (95% CI = 1·3–3·7%) for the dexamethasone arm compared with 5·0% (95% CI = 3·4–6·6%) for prednisolone (2P = 0·007 and see Fig 3). The overall CNS relapse rate (isolated and combined with relapse at another site) was also significantly less with dexamethasone (2P = 0·0004), as was the rate of relapse not involving the CNS (2P = 0·002). There was no significant difference in induction deaths or in deaths in remission between the prednisone and dexamethasone groups, so that EFS was significantly improved with dexamethasone (2P = 0·0007) (see Table V).

Figure 3.

Isolated central nervous system relapses by randomized steroid. Obs./Exp., observed/expected ratio.

Table V.  Outcome by steroid randomization.
 Prednisolone (n = 805)Dexamethasone (n = 798)O − EVOR (95%CI)2P
  1. Total numbers of events, and (in brackets) actuarial percentages at 5 years by randomized steroid allocation.

  2. O, observed; E, expected; V, variance; OR (95% CI), odds ratio with 95% confidence limits; 2P, double-sided P-value.

Isolated CNS relapse36 (5·0%)17 (2·5%)−9·7013·250·48 (0·28–0·82)0·007
No remission3 (0·4%)8 (1·0%)2·522·732·52 (0·77–8·24)0·1
Any CNS relapse64 (9·5%)31 (4·8%)−17·1823·740·48 (0·32–0·72)0·0004
Non-CNS relapse89 (13·6%)55 (6·9%)−18·7535·950·59 (0·43–0·82)0·002
Death in remission23 (3·0%)31 (4·1%)4·0013·501·34 (0·79–2·29)0·3
Any event179 (24·4%)125 (15·8%)−29·4175·930·68 (0·54–0·85)0·0007
Any death101 (14·2%)82 (11·0%)−9·4245·740·81 (0·61–1·09)0·2

At 5 years, the EFS was 84·2% (95% CI = 81·5–86·9%) with dexamethasone and 75·6% (95% CI = 72·3–78·9%) with prednisolone (Fig 4). OS was not significantly different between dexamethasone [89·0% (95% CI = 86·6%–91·3%)] and prednisolone [85·8% (95% CI = 83·1%–88·5%)] at 5 years (see Table IV). Analyses stratified by Thiopurine type and background treatment (ALL97, ALL97/99 regimens A, B or C) gave very similar results.

Figure 4.

Event-free survival by randomized steroid. Obs./Exp., observed/expected ratio.

There was no significant heterogeneity of treatment effect on EFS between subgroups by gender, age, WBC, immunophenotype or phase of trial. There was a suggestion that the relative risk reduction for CNS relapse with dexamethasone was greatest for those aged 10 years and above (P-value for heterogeneity = 0·03) and for non-CNS relapse for those aged under 10 years (P = 0·05) (see Fig 5).

Figure 5.

Forest plot showing the relative effect on EFS of steroid type within subgroups. The observed minus expected (O − E) number of events and its variance (Var.) are given for each subgroup. The odds ratio, calculated from these data, is shown as a filled box, with a line indicating the 95% confidence interval (CI). The overall effect after stratification by each subgroup variable is shown as a diamond whose width indicates the 95% CI of the overall result. The results of tests for heterogeneity, or trend in the case of WBC, are displayed to indicate any evidence of a different effect in any subgroup.

The numbers of early deaths – within 60 days of diagnosis – are shown in Table VI. There were no statistically significant differences between the steroid treatment groups whether analysed by age group or treatment arm.

Table VI.  Early deaths (<60 days) by randomized steroid, phase of trial (All97 or ALL99) and treatment regimen excluding ALLHR1 patients.
SteroidRand DRand to PNon-rand DNon-rand P
Early deaths14602
Age (years)<22–910+
SteroidRand DRand PRand DRand PRand DRand P
Early deaths0111431
SteroidRand DRand PRand DRand P
Early deaths11432
  1. There were no significant differences between treatment groups or age groups.

  2. Rand, randomized; D, dexamethasone; P, prednisolone.

 Initial regimenFinal regimen 
Treatment regimenABABC
Early deaths43223

Steroid toxicity

Table VII shows the relative incidence for each recorded toxicity by randomized steroid. There was a significant excess of overall toxicity in the dexamethasone group (11% vs. 5% with prednisolone). This consisted principally of an excess of behavioural problems, myopathy, osteopenia, and other toxicity (especially excess weight gain and acute liver enlargement). Treatment was changed from dexamethasone to prednisolone in 6% of patients because of unacceptable side effects.

Table VII.  Numbers of patients with grade III/IV steroid toxicity by randomized steroid.
 ALL97ALL99TotalRelative risk (95% CI)
Dexa (n = 389)Pred (n = 392)Dexa (n = 409)Pred (n = 413)Dexa (n = 798)Pred (n = 805)
  1. The relative risk of toxicity with dexamethasone compared with prednisolone was greater in ALL97 for behaviour (P = 0·02), any toxicity (P = 0·001), and any toxicity excluding behaviour (P = 0·01).

  2. *P < 0·05; **P < 0·001; ***P < 0·0001.

  3. †Numbers do not add up to total as some patients had more than one type of toxicity.

Behaviour***26221947114·31 (2·25–8·26)
Hypertension4024641·51 (0·43–5·35)
Diabetes6271013121·09 (0·50–2·38)
Myopathy**131932245·55 (1·92–16·04)
AVN1158690·67 (0·24–1·88)
Osteopenia*4130717·06 (0·87–57·18)
Other*2051717·07 (0·87–56·27)
Any†***437463289392·30 (1·60–3·31)
Any, excluding behaviour†*225282550301·68 (1·08–2·62)

Behavioural toxicity was reported in 6% of those allocated dexamethasone compared with 1% allocated prednisolone. These results must be treated with some caution, as the clinicians reporting them were not blind to treatment allocation. The behavioural changes ranged from mood swings and lability through to severe depression and violence towards self or others. There was a trend for girls to develop depression and for boys to be aggressive. No cases of suicide were recorded. However, three of the patients, all on dexamethasone, developed delusional psychoses. Cessation of steroids led to rapid resolution. Those patients on dexamethasone with severe toxicity were changed to prednisolone, with no reported significant recurrence of behavioural problems. Since analyses are by intention to treat, these patients’ outcomes are included in the dexamethasone group. No patient randomized to prednisolone was reported to have such severe mood change that they could not continue therapy with at least half dose of steroid.

Although seen in both arms of the study, there was a five fold higher incidence of myopathy on dexamethasone (2·8% vs. 0·5%). Almost all of these patients had lower limb involvement, most frequently of the quadriceps and/or glutei. Occasional distal calf and proximal upper limb weakness was reported although usually in conjunction with proximal lower limb weakness. For the vast majority of cases, recovery occurred after induction and the steroid type was not altered.

Severe osteopenia, rather than avascular necrosis (AVN), was rare but almost exclusively limited to patients receiving dexamethasone. This complication was associated with stress fractures of limbs and of the spine including laminar planar in four cases and severe bone pain in the remaining patients. A single patient on prednisolone developed bilateral stress fractures of the tibia having previously been changed from dexamethasone for hypertension and ‘difficult to control’ diabetes.

The most potentially long-term damaging effect of steroid therapy is AVN of the femoral head or tarsus tali. There is no excess of AVN in the dexamethasone arm at present. AVN was more frequent among older children, girls, and in the second phase of the trial. These factors were independently significant by logistic regression (P < 0·0001, 0·02, 0·03, respectively). After allowing for age, there was no effect of ALL99 schedule (A, B or C).

Once the Data Monitoring Committee had recommended trial closure, those patients still on treatment were changed to dexamethasone. We are now seeing late-onset behavioural problems and two cases so far of AVN in those patients randomized to prednisolone but who later received dexamethasone.

Diabetes was not more frequent with dexamethasone, but was more common at older ages with either steroid, with an incidence of 1·0% under 10 years, and 4·2% in patients aged 10 years or more (P < 0·0001).

No significant interaction between treatment and age was found for any individual toxicity. There was a significant interaction for all toxicities combined (P < 0·001), with increasing incidence of toxicity with age and no toxicity in patients under 2 years in the prednisolone arm. There was no such increase of toxicity with age in the dexamethasone group (Table VIII).

Table VIII.  Any steroid toxicity by age within randomized allocation.
 Dexamethasone [age (years)]Prednisolone [age (years)]
<2 (n = 61)2–9 (n = 607)≥10 (n = 130)<2 (n = 63)2–9 (n = 612)≥10 (n = 130)
Any steroid toxicity, n (%)10 (16%)62 (10%)17 (13%)0 (0%)19 (3%)20 (15%)


ALL97/99 demonstrated a considerable improvement in EFS compared with its predecessor UKALL XI (Hann et al, 2001), which showed no improvement over its predecessor, UKALL X (Chessells et al, 1995b). There are a variety of possible reasons why improvement has now occurred. Apart from the steroid randomization, there have been changes in the delivery and type of asparaginase, a major revision of the style of intensification and the introduction of a risk-stratified treatment regimen, which has led to further intensification of therapy for selected groups of patients.

The US CCG trial 1922, which closed in August 1995 after accruing 1062 standard risk patients (age 1–10 years of age, presenting WBC < 50 × 109/l) compared the efficacy of dexamethasone with prednisolone and also asked whether intravenous or oral 6-MP would carry greater advantage. No data were available for this study when ALL 97 opened, but subsequently it has been reported that the use of dexamethasone reduced the risk of central nervous system relapse so that the 6-year isolated CNS relapse rate was 3·7% (±0·8%) for those receiving dexamethasone and 7·1% (±1·1%) for the prednisolone arm (P = 0·01). There were also fewer isolated bone marrow relapses with dexamethasone, but the difference was not significant. The 6-year EFS was 85 ± 2% for the dexamethasone and 77 ± 2% for the prednisone arm (P = 0·002). The CCG trial showed comparable EFS for either route of administration for the 6-MP but, if the patients received intravenous therapy, retrieval after relapse was rarer (Bostrom et al, 2003).

The results for ALL97/99 were very similar to CCG1922, with reduction of both CNS and non-CNS relapse with dexamethasone (2·5% vs. 5·0% isolated CNS relapse at 5 years), resulting in improved EFS (84·2% vs. 75·6%). Although the CCG study was restricted to standard risk patients, our study included more high-risk patients (i.e. aged over 10 years or with presenting WBC > 50 × 109/l), suggesting that the benefit of treatment with dexamethasone in improving EFS also extends to high risk patients. Combining data from the UK and US trials to improve statistical power may indicate whether there are any subgroups in which the effect of using dexamethasone is increased or decreased. We hope to undertake a meta-analysis of the two trials in the near future.

The price of this improved outcome is, of course, greater steroid toxicity, with an increased incidence of behavioural problems, myopathy, overall osteopenia and excessive weight gain. AVN was not apparently increased with dexamethasone, but was more frequent among older children, females and in the second phase of the trial. Only longer-term follow-up will enable the true overall incidence of AVN to be determined. Diabetes was more frequent among older children (>10 years) but its incidence was not affected by the type of steroid administered. There was an interaction between treatment and age, with toxicity increasing with age in the prednisolone arm but not the dexamethasone arm, but it is not clear which specific toxicities definitely contribute to this interaction. Longer follow-up is needed to monitor toxicity, and, again data from the two large trials (UK and USA) will enable determination of whether there are specific groups at higher risk of side effects.

It has been suggested (Hurwitz et al, 2000; Silverman et al, 2001) that substituting dexamethasone for prednisolone complicates remission induction, and results in a higher incidence of septic episodes and death from toxicity. However, these studies were small and were not randomized for a steroid comparison. Although we initially had an excess of deaths in induction in patients receiving dexamethasone, this difference subsequently disappeared and, for the latter stages of the trial, the overall induction death rate was only five of 825 patients (0·6%) with no difference seen between the two steroid groups. Similarly, it has been suggested that the use of Daunorubicin in combination with dexamethasone during induction would lead to enhanced toxicity (Belgaumi et al, 2003). Our results do not support this contention. We saw only four early deaths in 577 patients in regimen A (no Daunorubicin), compared to only three early deaths in regimen B (which included Daunorubicin). There was no significant difference between these results (see Table IV).

We have successfully demonstrated that the use of dexamethasone in induction and subsequent phases of treatment for ALL results in a significant decrease in the risk of CNS relapse for all risk-groups of patients. Although this improvement in outcome was associated with an increase in toxicity, there was no excess of deaths attributable to the use of dexamethasone. This steroid, used at a dose in the order of 6·5 mg/m2 should now be regarded in future trials as part of standard therapy for childhood ALL. It remains open to question whether the improvement seen is because of some specific feature of dexamethasone or whether the equivalent dose to 6·5 mg/m2 of dexamethasone is greater than 40 mg/m2 or even higher, of prednisolone. At such doses, the toxicity profile for prednisolone could be quite different.


This clinical trial was supported by the Medical Research Council, with some additional support from Cancer Research UK.